Empathic accuracy was operationalized as the time-series correlation between a participant’s ratings of a target’s affect and the target’s own ratings of his or her affect. We calculated a separate correlation coefficient reflecting each participant’s overall empathic-accuracy score for each video clip (see Zaki et al., 2009, for details). We then modeled empathic accuracy as a function of drug condition (placebo= 0, oxytocin= 1), AQ (z score, continuous), and the Drug Condition× AQ interaction, using a mixed linear model. Because target expressivity is strongly associated with perceiver accuracy (Zaki et al., 2008), we included target expressivity (ie, targets’ scores on the Berkeley Expressivity Questionnaire; see Gross & John, 1997) as a covariate. Mixed-model analyses were performed using PROC MIXED in SAS 9.1 (SAS Institute, 2002), with restricted maximum likelihood to estimate parameters and the Kenward-Roger method to calculate appropriate degrees of freedom (Littell, Milliken, Stroup, Wolfinger, & Schabenberger, 2006).